Short-term thermal load prediction method and device, medium and equipment

Abstract

The invention discloses a short-term thermal load prediction method and device, a medium and equipment. The method comprises the steps of obtaining a historical data set and prediction feature data of a prediction day, performing fuzzy clustering processing on the historical data set to obtain a plurality of different types of data subsets, and calculating a clustering center of each data subset; performing similarity relationship calculation on the predicted feature data and the clustering center of each data subset, and taking the data subset similar to the predicted feature data as a training sample according to a calculation result; training the initial thermal load prediction model based on the training sample to obtain a trained thermal load prediction model; and inputting the prediction characteristic data into the trained thermal load prediction model to obtain thermal load prediction data of the prediction day, and training only for samples similar to meteorological parameters of the prediction day, so that thermal load prediction of the prediction day can be accurately performed, and the prediction precision and prediction efficiency of the short-term thermal load are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of heat load prediction, and in particular to a short-term heat load prediction method, device, medium and equipment. Background Art

[0002] Residential heating is a basic service industry related to people's livelihood and well-being, and is also an important part of social energy consumption. The ideal goal of residential heating is to achieve on-demand heating, that is, the heat source transmits heat to heat users through the primary network, heat exchange station and secondary network, and the heat changes with the demand of heat users. However, the heating load fluctuates greatly with meteorological conditions, and heat transmission has hysteresis. This requires the heating system to continuously optimize the operation strategy and make short-term predictions of the heat load based on meteorological information and historical characteristic parameters such as heat load, so that heating companies can finely regulate the heat load, thereby effectively improving the operating energy efficiency of the source side of the heating system and the satisfaction of the demand side.

[0003] At present, the heat load prediction method is mainly based on the neural network algorithm. The neural network algorithm is trained with a large number of samples for various characteristic parameters of various meteorological conditions, resulting in low prediction accuracy. Summary of the invention

[0004] In view of this, the present invention provides a short-term heat load prediction method, device, medium and equipment, the main purpose of which is to solve the problem of low accuracy of existing short-term heat load prediction methods when predicting heat load.

[0005] According to one aspect of the present application, a short-term heat load prediction method is provided, the method comprising:

[0006] Obtaining historical data sets and forecast characteristic data of the forecast day, performing fuzzy clustering processing on the historical data sets to obtain multiple data subsets of different categories, and calculating the cluster center of each data subset, wherein the data in the historical data sets and the forecast characteristic data of the forecast day are both between 0 and 1, and the historical data sets include actual characteristic data and actual heat load data of the data samples;

[0007] Calculating similarity between the predicted feature data and the cluster center of each data subset, and taking the data subset similar to the predicted feature data as a training sample according to the calculation result;

[0008] Training the initial heat load prediction model based on the training samples to obtain a trained heat load prediction model;

[0009] The predicted characteristic data is input into the trained heat load prediction model to obtain the heat load prediction data for the prediction day.

[0010] Optionally, the historical data set is subjected to fuzzy clustering processing to obtain a plurality of data subsets of different categories, including:

[0011] Based on the historical data set, a fuzzy similarity matrix corresponding to the historical data set is calculated, and a fuzzy equivalent matrix of the fuzzy similarity matrix is ​​obtained by using a square method;

[0012] Based on a plurality of preset threshold values ​​λ, the fuzzy equivalent matrix is ​​cut one by one to obtain a cut matrix corresponding to each threshold value λ, and data classification is performed based on each of the cut matrices to obtain a data classification result corresponding to each threshold value λ;

[0013] Based on multiple threshold values ​​λ and data classification results corresponding to the multiple threshold values ​​λ, a threshold association value corresponding to each of the threshold values ​​λ is calculated, and the data classification result under the threshold value λ corresponding to the maximum threshold association value is taken as the optimal classification result;

[0014] According to the optimal classification result and the historical data set, a plurality of data subsets of different categories are obtained.

[0015] Optionally, the calculating, based on the multiple threshold values ​​λ and the data classification results corresponding to the multiple threshold values ​​λ, a threshold association value corresponding to each threshold value λ includes:

[0016] According to the data classification results of the i-th cluster and the i-1-th cluster, the number of categories of the i-th cluster and the i-1-th cluster is obtained, and a first difference between the number of categories of the i-th cluster and the number of categories of the i-1-th cluster is calculated;

[0017] Calculate a second difference between the threshold λ of the i-1th cluster and the threshold λ of the i-th cluster;

[0018] The quotient of the first difference and the second difference is used as a threshold association value corresponding to a threshold λ of the i-th clustering.

[0019] Optionally, the following method is used to calculate the element in the i-th row and j-th column of the fuzzy similarity matrix:

[0020]

[0021] Among them, r ij is the similarity between the data of the i-th sample and the data of the j-th sample in the historical data set, r ij ∈[0,1], 0≤i≤n, 0≤j≤n; x ik ' is the kth data of the i-th sample, x jk ' is the kth data of the jth sample.

[0022] Optionally, the using the square method to obtain the fuzzy equivalent matrix of the fuzzy similarity matrix includes:

[0023] Obtain the square of the fuzzy similarity matrix in turn: R 2 =R·R, R 4 =R 2 ·R 2 , …, R 2k =R k ·R k Until R appears 2k =R k , fuzzy equivalence matrix R * =R k , where R is the fuzzy similarity matrix.

[0024] Optionally, based on the preset multiple thresholds λ, the fuzzy equivalent matrix is ​​cut one by one to obtain a cut matrix corresponding to each threshold λ, and data classification is performed based on each of the cut matrices to obtain a data classification result corresponding to each threshold λ, including:

[0025] The elements in the fuzzy equivalence matrix that are greater than or equal to a threshold value λ are replaced with 1, and the elements in the fuzzy equivalence matrix that are less than the threshold value λ are replaced with 0, to obtain a cut matrix;

[0026] The elements of each row in the cut matrix are compared, and rows with all the same elements from left to right are classified into one category to obtain the data classification result corresponding to the threshold λ.

[0027] Optionally, the performing similarity calculation on the predicted feature data and the cluster center of each data subset, and taking the data subset similar to the predicted feature data as a training sample according to the calculation result, includes:

[0028] The Euclidean distance between the predicted feature data and the cluster center of each data subset is calculated, and the data subset corresponding to the shortest Euclidean distance is used as a training sample.

[0029] According to another aspect of the present application, a short-term heat load prediction device is provided, comprising:

[0030] A fuzzy clustering processing module is used to obtain the historical data set and the predicted characteristic data of the prediction day, perform fuzzy clustering processing on the historical data set to obtain multiple data subsets of different categories, and calculate the cluster center of each data subset, wherein the data in the historical data set and the predicted characteristic data of the prediction day are both between 0 and 1, and the historical data set includes the actual characteristic data and actual heat load data of the data sample;

[0031] A training sample determination module is used to calculate the similarity relationship between the predicted feature data and the cluster center of each data subset, and according to the calculation result, use the data subset similar to the predicted feature data as a training sample;

[0032] A model training module, used to train the initial heat load prediction model based on the training samples to obtain a trained heat load prediction model;

[0033] The prediction data acquisition module is used to input the prediction feature data into the trained heat load prediction model to obtain the heat load prediction data of the prediction day.

[0034] Optionally, the fuzzy clustering processing module includes:

[0035] A matrix conversion unit, used for calculating the fuzzy similarity matrix corresponding to the historical data set based on the historical data set, and obtaining the fuzzy equivalent matrix of the fuzzy similarity matrix by using the square method;

[0036] A data classification unit, configured to cut the fuzzy equivalent matrix one by one based on a plurality of preset threshold values ​​λ to obtain a cut matrix corresponding to each threshold value λ, and to perform data classification based on each of the cut matrices to obtain a data classification result corresponding to each threshold value λ;

[0037] an optimal classification result determination unit, configured to calculate a threshold association value corresponding to each threshold λ based on multiple thresholds λ and data classification results corresponding to the multiple thresholds λ, and to take the data classification result under the threshold λ corresponding to the maximum threshold association value as the optimal classification result;

[0038] The data subset acquisition unit is used to obtain a plurality of data subsets of different categories according to the optimal classification result and the historical data set.

[0039] Optionally, the optimal classification result determination unit is further used for:

[0040] According to the data classification results of the i-th cluster and the i-1-th cluster, the number of categories of the i-th cluster and the i-1-th cluster is obtained, and a first difference between the number of categories of the i-th cluster and the number of categories of the i-1-th cluster is calculated;

[0041] Calculate a second difference between the threshold λ of the i-1th cluster and the threshold λ of the i-th cluster;

[0042] The quotient of the first difference and the second difference is used as a threshold association value corresponding to a threshold λ of the i-th clustering.

[0043] Optionally, the following method is used to calculate the element in the i-th row and j-th column of the fuzzy similarity matrix:

[0044]

[0045] Among them, r ijis the similarity between the data of the i-th sample and the data of the j-th sample in the historical data set, r ij ∈[0,1], 0≤i≤n, 0≤j≤n; x ik ' is the kth data of the i-th sample, x jk ' is the kth data of the jth sample.

[0046] Optionally, the matrix conversion unit is further used for:

[0047] Obtain the square of the fuzzy similarity matrix in turn: R 2 =R·R, R 4 =R 2 ·R 2 , …, R 2k =R k ·R k Until R appears 2k =R k , fuzzy equivalence matrix R * =R k , where R is the fuzzy similarity matrix.

[0048] Optionally, the data classification unit is further used for:

[0049] The elements in the fuzzy equivalence matrix that are greater than or equal to a threshold value λ are replaced with 1, and the elements in the fuzzy equivalence matrix that are less than the threshold value λ are replaced with 0, to obtain a cut matrix;

[0050] The elements of each row in the cut matrix are compared, and rows with all the same elements from left to right are classified into one category to obtain the data classification result corresponding to the threshold λ.

[0051] Optionally, the training sample determination module is used to:

[0052] The Euclidean distance between the predicted feature data and the cluster center of each data subset is calculated, and the data subset corresponding to the shortest Euclidean distance is used as a training sample.

[0053] According to another aspect of the present application, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned short-term heat load prediction method.

[0054] According to another aspect of the present application, a computer device is provided, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0055] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned short-term heat load prediction method.

[0056] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0057] The present application provides a short-term heat load prediction method, device, equipment and medium, which perform fuzzy clustering processing on historical data sets to obtain multiple data subsets of different categories, calculate the cluster center of each data subset, calculate the similarity between the predicted feature data and the cluster center of each data subset, and according to the calculation results, use the data subset most similar to the predicted feature data as a training sample, and train the initial heat load prediction model based on the training sample to obtain a trained heat load prediction model; input the predicted feature data of the prediction day into the trained heat load prediction model to obtain the heat load prediction data of the prediction day, and only train samples similar to the meteorological characteristic parameters of the prediction day, so as to accurately predict the heat load of the prediction day, improve the prediction accuracy and prediction efficiency of the heat load, and help to improve the stability and economy of the heating network.

[0058] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0060] Figure 1 A flow chart of a short-term heat load prediction method provided by an embodiment of the present application is shown;

[0061] Figure 2 A flow chart of another short-term heat load prediction method provided by an embodiment of the present application is shown;

[0062] Figure 3 A schematic diagram showing a heat load prediction model of a short-term heat load prediction method provided in an embodiment of the present application is shown;

[0063] Figure 4 A flow chart of another short-term heat load prediction method provided by an embodiment of the present application is shown;

[0064] Figure 5A structural block diagram of a short-term heat load prediction device provided in an embodiment of the present application is shown;

[0065] Figure 6 A schematic structural diagram of a computer device provided by an embodiment of the present invention is shown.

[0066] in,

[0067] Figure 5 In: 502- fuzzy clustering processing module; 504- training sample determination module; 506- model training module; 508- prediction data acquisition module;

[0068] Figure 6 In it: 602 - processor; 604 - communication interface; 606 - memory; 68 - communication bus; 610 - program. DETAILED DESCRIPTION

[0069] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0070] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention application are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0071] In view of the problem that the existing short-term heat load prediction method has low accuracy when predicting heat load, the embodiment of the present application provides a short-term heat load prediction method, such as Figure 1 As shown, the method includes:

[0072] 102: Obtain the historical data set and the predicted characteristic data of the prediction day, perform fuzzy clustering processing on the historical data set to obtain multiple data subsets of different categories, and calculate the cluster center of each data subset, wherein the data in the historical data set and the predicted characteristic data of the prediction day are both between 0 and 1, and the historical data set includes actual characteristic data and actual heat load data of the data sample;

[0073] 104: Calculate the similarity relationship between the predicted feature data and the cluster center of each data subset, and according to the calculation result, use the data subset similar to the predicted feature data as a training sample;

[0074] 106: training the initial heat load prediction model based on the training samples to obtain a trained heat load prediction model;

[0075] 108: Input the predicted feature data into the trained heat load prediction model to obtain the heat load prediction data for the prediction day.

[0076] Specifically, the historical heating load of the thermal power plant and the corresponding historical meteorological information, day type and other characteristic parameters are collected. The meteorological information includes the hourly maximum temperature, minimum temperature, wind speed, humidity and solar radiation intensity, etc. The day type parameters are set according to the demand for heating on working days, ordinary rest days, important rest days, etc. For example, important rest days are set to 1, ordinary rest days are set to 0.85, working days are set to 0.7, etc. The historical meteorological information and day type parameters are used as the initial historical data set, and the abnormal characteristic parameters in the initial historical data set are removed and replaced, and then the data in the data set are normalized so that the historical characteristic data is between 0-1, and the historical data set is obtained. The predicted characteristic data of the initial prediction day are also normalized so that the predicted characteristic data is also between 0-1. Perform fuzzy clustering on the historical data set to obtain multiple data subsets of different categories, calculate the cluster center of each data subset, calculate the Euclidean distance between the predicted feature data and the cluster center of each data subset, use the data subset corresponding to the shortest Euclidean distance as the most similar training sample, perform neural network training based on the most similar training sample, input the predicted feature data of the prediction day into the trained neural network, and obtain the predicted data of short-term heat load.

[0077] The present application provides a short-term heat load prediction method. Compared with the prior art, the present application performs fuzzy clustering processing on historical data sets to obtain multiple data subsets of different categories, calculates the cluster center of each data subset, calculates the similarity between the predicted feature data and the cluster center of each data subset, and according to the calculation results, uses the data subset most similar to the predicted feature data as a training sample, and trains an initial heat load prediction model based on the training sample and its corresponding heat load historical data to obtain a trained heat load prediction model; inputs the predicted feature data of the prediction day into the trained heat load prediction model to obtain the heat load prediction data of the prediction day, and only trains samples with similar meteorological parameters to the prediction day, so that the heat load prediction of the prediction day can be accurately performed, thereby improving the prediction accuracy and prediction efficiency of the heat load, and helping to improve the stability and economy of the heating network.

[0078] In one embodiment, hourly heat load, meteorological parameters (including hourly maximum temperature, minimum temperature, wind speed, humidity, solar radiation intensity, etc.), and day type are selected from the historical data of the thermal power plant as historical characteristic parameters. Specifically, the characteristic parameters of each sample are: heat load, maximum temperature, minimum temperature, wind speed, humidity, solar radiation intensity, and date type at the corresponding time of the day before and two days before the sample day, and maximum temperature, minimum temperature, wind speed, humidity, solar radiation intensity, and date type at the corresponding time of the sample day. Based on this, the initial sample set for short-term heat load prediction is:

[0079]

[0080] Where n is the number of samples; m is the number of historical characteristic parameters.

[0081] The initial sample set is preprocessed. First, the accuracy of the historical characteristic parameters in the initial sample set is verified, and the historical characteristic parameters of the samples are compared vertically and horizontally. Abnormal parameters are eliminated and replaced using the Lagrange interpolation method. Then, individual data in the sample set are normalized. If some characteristic data is in percentage form, no normalization is required.

[0082] The temperature in the characteristic parameters is normalized using the following formula:

[0083]

[0084] Where T is the temperature value and T' is the normalized temperature value.

[0085] The preprocessed historical data set is:

[0086]

[0087] The forecast feature data of the forecast day is preprocessed, and the feature parameter vector of the forecast day after preprocessing is:

[0088] X' p =[x' p1 ,x' p2 ,…,x' pm ]

[0089] In the formula, x' pj For X' p The j-th characteristic parameter of .

[0090] The heat load corresponding to each sample in the historical data set is normalized using the following formula:

[0091] y i '=(y i -y min ) / (y max -ymin )

[0092] In the formula, y i is x in the initial sample set X i 'The heat load value corresponding to the sample, y min ,y min are the maximum and minimum values ​​of the heat load values ​​corresponding to all samples, y i ' is the normalized heat load value.

[0093] In another embodiment of the present invention, in order to further define and illustrate, as Figure 2 As shown in the figure, fuzzy clustering is performed on the historical data set to obtain multiple data subsets of different categories, including:

[0094] 202: Based on the historical data set, calculate the fuzzy similarity matrix corresponding to the historical data set, and use the square method to find the fuzzy equivalent matrix of the fuzzy similarity matrix;

[0095] Specifically, based on the preprocessed historical data set X′, the absolute value index method is used to calculate the fuzzy similarity matrix R corresponding to the historical data set:

[0096]

[0097] The following method is used to calculate the element in the i-th row and j-th column of the fuzzy similarity matrix:

[0098]

[0099] Among them, r ij is the similarity between the data of the i-th sample and the data of the j-th sample in the historical data set, r ij ∈[0,1], 0≤i≤n, 0≤j≤n; x ik ' is the kth data of the i-th sample, x jk ' is the kth data of the jth sample.

[0100] The fuzzy equivalent matrix of the fuzzy similarity matrix is ​​obtained by using the square method, including:

[0101] Calculate the square of the fuzzy similarity matrix in turn: R 2 =R·R, R 4 =R 2 ·R 2 , …, R 2k =R k ·R k , until a natural number k (k≤n) appears such that R 2k =R k , fuzzy equivalence matrix R * =R k, where R is the fuzzy similarity matrix.

[0102] 204: based on a plurality of preset threshold values ​​λ, the fuzzy equivalent matrix is ​​cut one by one to obtain a cut matrix corresponding to each threshold value λ, and data classification is performed based on each cut matrix to obtain a data classification result corresponding to each threshold value λ;

[0103] Specifically, for a cut of the fuzzy equivalence matrix, the elements in the fuzzy equivalence matrix that are greater than or equal to the threshold λ are replaced with 1, and the elements in the fuzzy equivalence matrix that are less than the threshold λ are replaced with 0 to obtain a cut matrix; the elements in each row of the cut matrix are compared, and the rows with the same elements in the same column are grouped into one category to obtain the data classification result corresponding to the threshold λ.

[0104] The above method is used to cut the fuzzy equivalence matrix based on each threshold λ, and the data classification result corresponding to each threshold λ is obtained.

[0105] 206: Calculate a threshold correlation value corresponding to each threshold λ based on the multiple thresholds λ and the data classification results corresponding to the multiple thresholds λ, and use the data classification result under the threshold λ corresponding to the maximum threshold correlation value as the optimal classification result;

[0106] Specifically, according to the data classification results of the i-th cluster and the i-1-th cluster, the number of categories of the i-th cluster and the i-1-th cluster is obtained, and the first difference between the number of categories of the i-th cluster and the number of categories of the i-1-th cluster is calculated; the second difference between the threshold λ of the i-1-th cluster and the threshold λ of the i-th cluster is calculated; the quotient of the first difference and the second difference is taken as the threshold association value corresponding to the threshold λ of the i-th cluster, and the data classification result under the threshold λ corresponding to the maximum threshold association value is taken as the optimal classification result.

[0107] The selection of the optimal threshold λ can be obtained by the following formula: i The maximum λ i The value is the optimal threshold λ.

[0108]

[0109] Where i represents the clustering order of threshold λ from high to low; n i and n i-1 Respectively represent the number of elements in the i-th cluster and the i-1-th cluster; λ i and λ i-1 They represent the thresholds of the i-th cluster and the i-1-th cluster respectively.

[0110] 208: Based on the optimal classification result and the historical data set, multiple data subsets of different categories are obtained.

[0111] According to the optimal classification result, the historical data corresponding to each category is obtained from the historical data set to generate data subsets of different categories.

[0112] In one embodiment, after fuzzy clustering is performed on the preprocessed historical data set, the k-th data subset is:

[0113]

[0114] The cluster center of the k-th data subset is:

[0115]

[0116] In the formula, is the jth characteristic parameter of the cluster center of the kth data subset, calculated using the following formula:

[0117]

[0118] In the formula, x i ' j represents the jth characteristic parameter of the i-th sample in the k-th data subset; n k is the number of samples in the k-th data subset, 0≤i≤n k , 0≤j≤m.

[0119] In one embodiment, similarity calculation is performed between the predicted feature data and the cluster center of each data subset, and based on the calculation result, the data subset similar to the predicted feature data is used as a training sample, including:

[0120] The Euclidean distance between the predicted feature data and the cluster center of each data subset is calculated, and the data subset corresponding to the shortest Euclidean distance is used as a training sample.

[0121] Specifically, the following formula is used to calculate the cluster center of each data subset and the predicted feature data X' p The Euclidean distance is used to select the data subset with the shortest distance as the training sample of the BP neural network.

[0122]

[0123] Where, d pk is the predicted day characteristic parameter vector X' p The Euclidean distance between the cluster center of the k-th data subset, x' pj is the jth data in the feature data of the forecast day, m is the number of feature data of the forecast day, is the jth data of the cluster center of the kth data subset.

[0124] In one embodiment, the heat load prediction model adopts a three-layer BP neural network including an input layer, a hidden layer, and an output layer, and adopts a variable learning rate η and a variation factor α to increase the convergence speed.

[0125] Specifically, Figure 3 As shown in the figure, the activation functions of the input layer and output layer of the BP neural network both use linear functions, namely:

[0126] f1(x)=f3(x)=x

[0127] The hidden layer activation function of the BP neural network adopts the hyperbolic tangent function, that is:

[0128]

[0129] Specifically, the number of input layer nodes is the number of sample feature parameters m, and the number of output layer nodes is 1; the number of hidden layer nodes can be obtained using the following formula:

[0130]

[0131] In the formula, h is the number of nodes in the hidden layer, m is the number of nodes in the input layer, and s is the number of nodes in the output layer.

[0132] The input vector X of the neuron in the i-th layer i And the output vector Y i They are calculated by the following formulas:

[0133] X i =W i ·Y i-1 +B i

[0134] Where W i , B i They are the connection weight matrix and bias vector between the i-1th neuron output layer and the i-th neuron input layer respectively; for the first layer, X1=W1·X0+B0.

[0135] Y i =f i (X i )

[0136] In the formula, f i (x) is the activation function of the i-th layer.

[0137] The training samples are forward propagated through each layer through the input layer, and the errors are back-propagated to update the weights and biases of the network to train the neural network.

[0138] In one embodiment, Figure 4As shown, the historical sample data is first preprocessed, and then fuzzy clustering is performed on the preprocessed historical sample data to obtain sample subsets of multiple categories, and the sample subsets most similar to the predicted day characteristics are found as training samples for the BP neural network prediction module. The BP neural network is trained based on the training samples to obtain a trained neural network, and the short-term heat load value is predicted based on the predicted day characteristic parameters and the trained neural network.

[0139] Furthermore, as a response to the above Figure 1 The embodiment of the present invention provides a short-term heat load prediction device, such as Figure 5 As shown, the device comprises:

[0140] The fuzzy clustering processing module 502 is used to obtain the historical data set and the predicted characteristic data of the prediction day, perform fuzzy clustering processing on the historical data set to obtain multiple data subsets of different categories, and calculate the cluster center of each data subset, wherein the data in the historical data set and the predicted characteristic data of the prediction day are both between 0 and 1, and the historical data set includes the actual characteristic data and actual heat load data of the data sample;

[0141] The training sample determination module 504 is used to calculate the similarity relationship between the predicted feature data and the cluster center of each data subset, and according to the calculation result, the data subset similar to the predicted feature data is used as a training sample;

[0142] A model training module 506 is used to train the initial heat load prediction model based on the training samples to obtain a trained heat load prediction model;

[0143] The prediction data acquisition module 508 is used to input the prediction feature data into the trained heat load prediction model to obtain the heat load prediction data of the prediction day.

[0144] The present application provides a short-term heat load prediction device. Compared with the prior art, the present application performs fuzzy clustering processing on historical data sets to obtain multiple data subsets of different categories, calculates the cluster center of each data subset, calculates the similarity between the predicted feature data and the cluster center of each data subset, and according to the calculation result, uses the data subset most similar to the predicted feature data as a training sample, and trains an initial heat load prediction model based on the training sample and its corresponding heat load historical data to obtain a trained heat load prediction model; inputs the predicted feature data of the prediction day into the trained heat load prediction model to obtain the heat load prediction data of the prediction day, and only trains samples with meteorological parameters similar to the prediction day, so that the heat load prediction of the prediction day can be accurately performed, thereby improving the prediction accuracy and prediction efficiency of the heat load, and helping to improve the stability and economy of the heating network.

[0145] In one embodiment, the fuzzy clustering processing module includes:

[0146] A matrix conversion unit, used for calculating the fuzzy similarity matrix corresponding to the historical data set based on the historical data set, and obtaining the fuzzy equivalent matrix of the fuzzy similarity matrix by using the square method;

[0147] A data classification unit is used to cut the fuzzy equivalent matrix one by one based on a plurality of preset threshold values ​​λ to obtain a cutting matrix corresponding to each threshold value λ, and to classify data based on each cutting matrix to obtain a data classification result corresponding to each threshold value λ;

[0148] An optimal classification result determination unit, configured to calculate a threshold association value corresponding to each threshold λ based on multiple thresholds λ and data classification results corresponding to the multiple thresholds λ, and take the data classification result under the threshold λ corresponding to the maximum threshold association value as the optimal classification result;

[0149] The data subset acquisition unit is used to obtain multiple data subsets of different categories based on the optimal classification result and the historical data set.

[0150] In one embodiment, the optimal classification result determination unit is further configured to:

[0151] According to the data classification results of the i-th cluster and the i-1-th cluster, the number of categories of the i-th cluster and the i-1-th cluster is obtained, and the first difference between the number of categories of the i-th cluster and the number of categories of the i-1-th cluster is calculated;

[0152] Calculate the second difference between the threshold λ of the i-1th cluster and the threshold λ of the i-th cluster;

[0153] The quotient of the first difference value and the second difference value is used as the threshold association value corresponding to the threshold λ of the i-th clustering.

[0154] In one embodiment, the following method is used to calculate the element in the i-th row and j-th column of the fuzzy similarity matrix:

[0155]

[0156] Among them, r ij is the similarity between the data of the i-th sample and the data of the j-th sample in the historical data set, r ij ∈[0,1], 0≤i≤n, 0≤j≤n; x ik ' is the kth data of the i-th sample, x jk ' is the kth data of the jth sample.

[0157] In one embodiment, the matrix conversion unit is further configured to:

[0158] Calculate the square of the fuzzy similarity matrix in turn: R2 =R·R, R 4 =R 2 ·R 2 , …, R 2k =R k ·R k Until R appears 2k =R k , fuzzy equivalence matrix R * =R k , where R is the fuzzy similarity matrix.

[0159] In one embodiment, the data classification unit is further configured to:

[0160] The elements in the fuzzy equivalence matrix that are greater than or equal to the threshold λ are replaced with 1, and the elements in the fuzzy equivalence matrix that are less than the threshold λ are replaced with 0, and a cut matrix is ​​obtained;

[0161] Compare the elements of each row in the cut matrix, and classify the rows with the same elements from left to right into one category to obtain the data classification result corresponding to the threshold λ.

[0162] In one embodiment, the training sample determination module is used to:

[0163] The Euclidean distance between the predicted feature data and the cluster center of each data subset is calculated, and the data subset corresponding to the shortest Euclidean distance is used as a training sample.

[0164] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the short-term heat load prediction method in any of the above method embodiments.

[0165] Figure 6 A schematic diagram of the structure of a computer device provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computer device.

[0166] like Figure 6 As shown, the computer device may include: a processor (processor) 602 , a communication interface (Communications Interface) 604 , a memory (memory) 606 , and a communication bus 608 .

[0167] The processor 602 , the communication interface 604 , and the memory 606 communicate with each other via a communication bus 608 .

[0168] The communication interface 604 is used to communicate with other devices such as clients or other servers.

[0169] The processor 602 is used to execute the program 610, and specifically can execute the relevant steps in the above-mentioned short-term heat load prediction method embodiment.

[0170] Specifically, the program 610 may include program codes, which include computer operation instructions.

[0171] The processor 602 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computer device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0172] The memory 606 is used to store the program 610. The memory 606 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0173] The program 610 may be specifically configured to enable the processor 602 to perform the following operations:

[0174] Obtain the historical data set and the predicted characteristic data of the forecast day, perform fuzzy clustering on the historical data set to obtain multiple data subsets of different categories, and calculate the cluster center of each data subset, wherein the data in the historical data set and the predicted characteristic data of the forecast day are both between 0 and 1, and the historical data set includes the actual characteristic data and actual heat load data of the data sample;

[0175] Calculate the similarity between the predicted feature data and the cluster center of each data subset, and use the data subset similar to the predicted feature data as a training sample based on the calculation results;

[0176] The initial heat load prediction model is trained based on the training samples to obtain a trained heat load prediction model;

[0177] The predicted feature data is input into the trained heat load prediction model to obtain the heat load prediction data for the prediction day.

[0178] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices. In one embodiment, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0179] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A short-term heat load prediction method, characterized in that: include: Obtaining historical data sets and forecast characteristic data of the forecast day, performing fuzzy clustering processing on the historical data sets to obtain multiple data subsets of different categories, and calculating the cluster center of each data subset, wherein the data in the historical data sets and the forecast characteristic data of the forecast day are both between 0 and 1, and the historical data sets include actual characteristic data and actual heat load data of the data samples; Calculating similarity between the predicted feature data and the cluster center of each data subset, and taking the data subset similar to the predicted feature data as a training sample according to the calculation result; Training the initial heat load prediction model based on the training samples to obtain a trained heat load prediction model; The predicted characteristic data is input into the trained heat load prediction model to obtain the heat load prediction data for the prediction day.

2. The short-term heat load prediction method according to claim 1, characterized in that: The fuzzy clustering process is performed on the historical data set to obtain multiple data subsets of different categories, including: Based on the historical data set, a fuzzy similarity matrix corresponding to the historical data set is calculated, and a fuzzy equivalent matrix of the fuzzy similarity matrix is ​​obtained by using a square method; Based on a plurality of preset threshold values ​​λ, the fuzzy equivalent matrix is ​​cut one by one to obtain a cut matrix corresponding to each threshold value λ, and data classification is performed based on each of the cut matrices to obtain a data classification result corresponding to each threshold value λ; Based on multiple threshold values ​​λ and data classification results corresponding to the multiple threshold values ​​λ, a threshold association value corresponding to each of the threshold values ​​λ is calculated, and the data classification result under the threshold value λ corresponding to the maximum threshold association value is taken as the optimal classification result; According to the optimal classification result and the historical data set, a plurality of data subsets of different categories are obtained.

3. The short-term heat load prediction method according to claim 2, characterized in that: The step of calculating a threshold association value corresponding to each threshold λ based on the multiple thresholds λ and the data classification results corresponding to the multiple thresholds λ includes: According to the data classification results of the i-th cluster and the i-1-th cluster, the number of categories of the i-th cluster and the i-1-th cluster is obtained, and a first difference between the number of categories of the i-th cluster and the number of categories of the i-1-th cluster is calculated; Calculate a second difference between the threshold λ of the i-1th cluster and the threshold λ of the i-th cluster; The quotient of the first difference and the second difference is used as a threshold association value corresponding to a threshold λ of the i-th clustering.

4. The short-term heat load prediction method according to claim 2, characterized in that: The following method is used to calculate the element in the i-th row and j-th column of the fuzzy similarity matrix: Among them, r ij is the similarity between the data of the i-th sample and the data of the j-th sample in the historical data set, r ij ∈[0,1], 0≤i≤n, 0≤j≤n; x ik ' is the kth data of the i-th sample, x jk ' is the kth data of the jth sample.

5. The short-term heat load prediction method according to claim 2, characterized in that: The method of using the square method to obtain the fuzzy equivalent matrix of the fuzzy similarity matrix includes: Obtain the square of the fuzzy similarity matrix in turn: R 2 =R·R, R 4 =R 2 ·R 2 , …, R 2k =R k ·R k Until R appears 2k =R k , fuzzy equivalence matrix R * =R k , where R is the fuzzy similarity matrix.

6. The short-term heat load prediction method according to claim 2, characterized in that: The method of cutting the fuzzy equivalent matrix based on the preset multiple thresholds λ one by one to obtain a cutting matrix corresponding to each threshold λ, and performing data classification based on each cutting matrix to obtain a data classification result corresponding to each threshold λ includes: The elements in the fuzzy equivalence matrix that are greater than or equal to a threshold value λ are replaced with 1, and the elements in the fuzzy equivalence matrix that are less than the threshold value λ are replaced with 0, to obtain a cut matrix; The elements of each row in the cut matrix are compared, and rows with all the same elements from left to right are classified into one category to obtain the data classification result corresponding to the threshold λ.

7. The short-term heat load prediction method according to claim 1, characterized in that: The similarity calculation is performed on the predicted feature data and the cluster center of each data subset, and according to the calculation result, the data subset similar to the predicted feature data is used as a training sample, including: The Euclidean distance between the predicted feature data and the cluster center of each data subset is calculated, and the data subset corresponding to the shortest Euclidean distance is used as a training sample.

8. A short-term heat load prediction device, characterized in that: include: A fuzzy clustering processing module is used to obtain the historical data set and the predicted characteristic data of the prediction day, perform fuzzy clustering processing on the historical data set to obtain multiple data subsets of different categories, and calculate the cluster center of each data subset, wherein the data in the historical data set and the predicted characteristic data of the prediction day are both between 0 and 1, and the historical data set includes the actual characteristic data and actual heat load data of the data sample; A training sample determination module is used to calculate the similarity relationship between the predicted feature data and the cluster center of each data subset, and according to the calculation result, use the data subset similar to the predicted feature data as a training sample; A model training module, used to train the initial heat load prediction model based on the training samples to obtain a trained heat load prediction model; The prediction data acquisition module is used to input the prediction feature data into the trained heat load prediction model to obtain the heat load prediction data of the prediction day.

9. A storage medium, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the short-term heat load prediction method according to any one of claims 1 to 8.

10. A computer device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the short-term heat load prediction method according to any one of claims 1-8.

Images